【摘要】：為得到更加精確的聲道共鳴特性,人們?cè)絹碓疥P(guān)心說話時(shí)聲道的結(jié)構(gòu)及其形狀變化。該文從MRI 3-D圖像數(shù)據(jù)中提取了漢語普通話7個(gè)單元音[a]、[o]、[酇]、[i]、[u]、[y]、[e]的聲道邊緣、中線和矢量半徑數(shù)據(jù),并沿著聲道中線從唇到聲門等間距截取聲道的36個(gè)截面,對(duì)于每個(gè)截面,根據(jù)矢量半徑的大小使同一個(gè)位置的截面形狀作線性過渡,從而建立用矢量半徑驅(qū)動(dòng)的漢語普通話立體聲道模型。計(jì)算模型的共振峰并合成語音樣本,在與自然語音對(duì)比的聽辨實(shí)驗(yàn)中,該模型取得了較好的語音合成效果。
[Abstract]:In order to obtain more accurate acoustic resonance, people pay more and more attention to the structure and shape change of the channel. In this paper, the data of channel edge, midline and vector radius of 7 unit sounds [a], [o], [Hu], [I], [u], [y], [e] of Chinese mandarin are extracted from MRI 3-D image data. At the same time, 36 sections of the vocal channel are intercepted along the midline of the channel from lip to glottis. For each section, the shape of the section at the same position is linearly transitioned according to the size of the vector radius. Thus, a stereo channel model driven by vector radius is established. The resonance peak of the model is calculated and the speech samples are synthesized. In the audiological experiment compared with the natural speech, the model achieves a good effect of speech synthesis.
【作者單位】： 河南大學(xué)文學(xué)院;北京大學(xué)中國語言文學(xué)系;
【基金】：中國哲學(xué)社會(huì)科學(xué)重大投標(biāo)項(xiàng)目(10&ZD125)
【分類號(hào)】：H017
，

本文編號(hào)：2349012